You type a task in plain English, "find a hotel near this address under $200 with free cancellation and add it to my cart," and the browser does it. It scrolls, clicks, fills forms, compares options, applies discount codes, and stops to ask you before it pays.

It is called auto browse. It runs in a side panel while you work on something else. It is live right now on Windows, Mac, and Chromebook in the US for anyone on Google AI Pro or Ultra, and it reaches Android at the end of June (source: 9to5Google).

This goes beyond one browser feature. The web itself is being rebuilt so agents can use it directly, not by guessing at screenshots but through a real standard. That shift is happening this year. The people who understand it first get hours back every week while everyone else still fills forms by hand.

This week: how to set up an agent that runs your repetitive web work, and where the agentic web is heading before December.

The Old Way:

Think about the web work you do that takes zero thought but real time. Booking the same kind of appointment. Copying data from a supplier portal into a spreadsheet. Checking five sites for one price. Filling the same form fields again and again. Filing an expense report. Comparing flights against a set of rules you hold in your head.

None of it is hard. All of it is slow. Knowledge workers lose around a quarter of the workday to this kind of repetitive digital chore, roughly a full day a week (source: BusinessPlusAI). You do not notice it because it comes in ten-minute slices spread across the week. Added up, it is the most expensive habit you have, and you are the one doing it.

The Replacement:

A browser agent that does the chore while you do the work that matters.

Auto browse, powered by Gemini 3 inside Chrome, takes a plain-English task and carries it out across the live web. It reads what is on the page, searches, compares, fills forms (including pulling details straight from a PDF), and builds a cart within a budget, even applying discount codes. With your permission it uses Google Password Manager to get through sign-ins. It pauses and asks before anything sensitive, like a purchase or a post (source: Google, 9to5Google).

The honest caveat, up front: auto browse is US-only right now and needs Google AI Pro ($25/month) or Ultra (source: Gemini Apps Help). If you do not have it, the Pro section includes a second build using Claude that any reader can run today. You are not locked out.

What this replaces: the day a week you lose to web chores.

What it costs: a subscription you may already have, or the Claude alternative.

Setup time: 15 minutes to configure. After that, every chore is one instruction.

Two years ago this did not exist in any usable form. The reason it works now is that the model behind it (Gemini 3) can look at a page the way you do, understand what each element is for, and act on it, instead of following a brittle script that breaks the moment a button moves.

That is also why it pauses before it pays. The agent is capable of completing a purchase end to end, but Google built in a hard stop for anything sensitive: payments, posts, and other actions you cannot undo (source: 9to5Google). You stay in control of the irreversible steps. The agent handles everything leading up to them.

The practical effect is that you stop being the one clicking through tabs. You describe the outcome, review the plan it proposes, and let it run. The chores that used to fragment your day get handled in the background while you do the work only you can do.

The Pro section below includes:

What you get: the repetitive web work off your plate, handled by an agent you brief once.

The Browser Agent Setup: The full 15-minute build, the settings that make auto browse fast instead of timid, 8 copy-paste task recipes (booking, data entry, price monitoring, forms, expenses, research pulls), and the security rules that keep it safe. Plus a Claude-based build for anyone without auto browse. → Complete walkthrough, copy-paste recipes, works whether or not you have Gemini Pro

The Agentic Web Playbook: How to make your website and your daily tools agent-ready before this becomes table stakes, the new web standard explained in plain English, and the three moves that put you ahead of competitors who have not noticed the shift. → Full breakdown with specific actions for your business

Monthly resource drops:

"I set up one automation from the toolkit last weekend and saved 6 hours in the first week alone. It feels like having an extra team member running in the background."

Elena M. DTC Founder.

🔒 Pro members get the full agent setup, all 8 task recipes, and the security rules, and have it running today. The build works whether or not you have Gemini Pro.

Zoom launched ZoomMate on June 1 at $20 per user per month. It sits inside your live meetings and connects what gets decided to what happens next across the tools where your work actually lives: Salesforce, Jira, ServiceNow, and Slack (source: AI Apps).

The feature worth the price is "Complete." It turns the raw notes from a meeting into a finished document or presentation on its own, with no blank-page step afterward. As Zoom's chief product officer put it, the tool connects what was decided to what needs to happen next across every system where your work lives (source: AI Apps).

Why this week: most meeting AI stops at a transcript and a summary, which still leaves you to do the actual follow-up. ZoomMate closes that gap by pushing the decisions into your other systems and drafting the deliverable. If meetings eat your calendar and the follow-up eats your evenings, this is the one to test.

🔒 Implementation checklist is for pro members. Get a breakdown plan to implement AI into your workflow.

🌐 Auto browse reaches Android at the end of June. Gemini in Chrome with auto browse rolls out to US Android devices with 4GB+ RAM for AI Pro and Ultra subscribers (source: 9to5Google). Why this matters for you: the agentic browser leaves the desktop and goes in your pocket. The chores you hand it can happen anywhere.

🔌 Google's WebMCP enters its Chrome origin trial. The standard that lets websites expose real tools to AI agents, instead of making them guess at screenshots, starts its origin trial in Chrome 149 (source: Google Developers Blog). Why this matters for you: this is the plumbing of the agentic web. It is the Big Story below.

⚡ Gemini 3.5 Flash beats last year's Pro tier. A Flash-tier model that outperforms the previous Pro on coding, reasoning, and multimodal, at Flash speed and cost (source: AI Made Tools). Why this matters for you: the cheap, fast model is now also the good one. For high-volume agent work, your cost just dropped.

🔬 Google opens experimental science agents. Gemini for Science put agentic peer review and hypothesis tools into pilots, and its Paper Assistant reviewed over 10,000 papers (source: Google Research). Why this matters for you: agents are moving into skilled expert work, not just chores. The template for automating professional workflows is forming now.

For two years, browser agents have worked like someone navigating a room blindfolded. Take a screenshot, guess what is clickable, click, wait for the page to change, screenshot again, guess again. It is slow, it breaks when a button moves two pixels, and it wastes tokens (source: byteiota, AgentMarketCap).

That era is ending. At I/O 2026, Google pushed WebMCP into a Chrome origin trial. WebMCP is a proposed open standard that lets a website expose structured tools, actual functions an agent can call, instead of forcing it to scrape the page and hope. Early measurements put the token saving at up to 89%.

This is not one company's play. It is a W3C draft. Microsoft Edge has added support. The realistic path is a cross-browser baseline within 18 to 24 months. Every major agent, Google's Gemini, OpenAI's Atlas, Anthropic's Claude, Perplexity's Comet, is racing to operate in this new layer.

The shift in plain terms: the web stops being only pages humans read and becomes a set of services that humans and agents both use. When that happens, how people get things done online changes. Fewer people clicking through five tabs. More people handing a goal to an agent and getting the result. The businesses on the receiving end of that traffic will be sorted into two groups: the ones agents can use, and the ones agents skip.

🔒 The full playbook is for pro members: What the agentic web looks like by December, the three moves to make your business agent-ready before it is table stakes, and why the sites that prepare now get recommended by AI assistants while the rest get skipped.

New This Week:

NEW Resources This Month:

Each resource lives permanently in your Pro account. Use them whenever you need them.

The web you use every day is being rebuilt under you. For the next few months, running an agent for your chores is an edge. By the end of the year it is normal. The people who set it up now get the head start, in time saved and in understanding where this goes before their competitors do.

If you do one thing before next Thursday, hand your most-dreaded web chore to an agent and watch it work. The first time it books the thing or fills the form on its own, you will start seeing your week differently.

👇 What’s behind the paywall:

Pro members deploy AI in their work an average of 5-8x more often than free readers (based on reply data from past issues). The difference is having the exact setup, not the concept.